Machine learning-based automated planning for hippocampal avoidance prophylactic cranial irradiation

Area of Science:

• Radiation Oncology
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Prophylactic cranial irradiation (PCI) is used in small-cell lung cancer treatment.
• Hippocampal avoidance (HA) is crucial to mitigate neurocognitive side effects.
• Automated treatment planning can improve efficiency and consistency.

Purpose of the Study:

• To design and evaluate a knowledge-based model for automated hippocampal avoidance prophylactic cranial irradiation (HA-PCI) planning.
• To assess the efficiency and clinical acceptability of AI-generated HA-PCI plans.
• To compare automated plans with manually generated plans.

Main Methods:

• A knowledge-based (KB) planning routine was developed using commercial AI tools.
• Data from 44 patients were used for training, with a second version validated on 9 additional plans.
• Automated plans (AP) were compared to manual plans (MP) using PREMER trial objectives for 10 test patients.

Main Results:

• Automated planning reduced optimization time by 55% (15 vs 33 min).
• AP showed statistically significant improvements in D98%, Homogeneity Index, and Hippocampus Dmean compared to MP.
• AP met objectives without deviation, while MP had deviations in 2 cases.

Conclusions:

• A knowledge-based model enables automated HA-PCI planning.
• Automation of radiotherapy planning enhances efficiency, safety, and quality.
• This approach may improve access to advanced radiotherapy techniques.